21st Century Geographical Methods for Timeless Spatial Inequality Challenges

Rachel Franklin (@rsfrankl)

Newcastle University | The Alan Turing Institute

A Tale Told in Three Chapters

Acknowledgements

Art Getis, Waldo Tobler, and Bill Clark

Prologue

What is spatial inequality?

The observation of uneven attributes or outcomes over space, for sure. But also the processes and structures producing that unevenness, as well as the impacts of those inequalities.

Chapter 1

Spatial inequality and the smart city

Thinking about how emerging technologies intersect with the spatial demography of cities to exacerbate, reproduce, and generate inequalities across areas or groups.

Especially placement of sensors in the urban landscape

  • Sensor coverage and “sensor deserts”
  • Who’s in the “gaps”
  • Equitable decision-making

Why should we care about sensors and sensor networks?

1. They’re everywhere

2. Increasingly they’re essential urban infrastructure

  • Policy decisions are made from sensor measurements
  • Sensors and sensor networks help cities run

3. We humans are funny about infrastructure and equity

4. Even with good intentions we miss people and places

The big idea

How can we support informed and equitable decision-making around sensor placement, especially what criteria ideal networks might satisfy and the inevitable trade-offs involved?

Geographers have methods for this! (yay us!)

  • What is the ”best” allocation of n sensors, given a particular goal?

    • Single objective greedy algorithm: Place sensors one by one and maximize coverage of one sub-group only.
    • Multi-objective genetic algorithm (NSGA2): Generate a spectrum of networks representing the coverage trade-offs between different sub-groups.
  • Decision support tools that visualize options and trade-offs

Coverage for older residents (>65)

Coverage for place-of-work population

Making it more user-friendly

No fancy data here and well-known methods



But should make us think:

  1. how are our data produced?

  2. how can we do better?

Chapter 2

Public transportation heat exposure in a warming world

Thinking about the ways in which health, climate change, and transportation intersect

Case in point: The London Tube

  • The first Tube line opened in 1863 and is still running as part of the Metropolitan line
  • 272 stations
  • 11 lines, covering 402 kilometres
  • More than five million passengers on the busiest days

Future heat on the London Tube

  • Only 4 out of 11 London underground lines have air conditioning systems

  • The average summer temperature in London is expected to increase by 2.7 degrees Celsius by the 2050s

  • The probability of heatwaves could also increase five-fold and they’re expected to occur every other year

  • By 2070, the mean maximum air temperature in the UK in August is projected to increase by up to 6 °C in summer compared to 2018

Hot is already here: Average station temperatures in London (2019)

The big idea

How can we estimate current and future heat exposure on the Tube and who (where) is most affected?

Simple (and important) question with complex data requirements

  • Travel flows (origins and destinations)

  • Who’s travelling? (demographic and health characteristics)

  • What’s the temperature on board? (estimated from known station-surface differentials)

Exciting geographical tools

  • Synthetic Population Catalyst (SPC)–A synthetic population that simulates individual-level travel behaviour (homeplace and workplace), travel mode, person and socio-economic factors that allow us to explore heat vulnerability at the individual level

  • Tube operation timetable–For travel times and route estimation, we use the timetable provided by TfL APIs. Provides accurate estimation whether travellers for each OD-pair will take air-conditioned Tube lines.

  • Clim-recal–Estimates weather and heat wave days in the past and future on a daily basis in 2.2 km*2.2 km cells covering the entire UK. Local variation in the dataset is used to estimate heat exposure more accurately.

Preliminary results (a sense of where we’re heading)

  • The inequality of heat exposure risk is significant in spatial terms

  • Trickiness of estimation but lots of useful data that can be brought to bear

  • But also some of this should be being measured directly!

Chapter 3

Making satellite imagery data usable, useful, and used in the social sciences and health

Thinking about how many decades1 we’ve been talking about the potential for satellite imagery to be a social science data game-changer.



What’s stopping us?

  • Lack of useful data products

  • Skills and capacity

  • Data products in formats (and locations) social scientists and health folks are used to

  • Interfaces that serve data that people want to use

  • Until pretty recently: high-resolution imagery, plus the tools to ingest, extract, and package at scale

The big idea

We’ve now got the compute, methods, and sensor quality for satellite imagery to be a game-changer for social science and health research and policy making

Introducing the ESRC SDR Imagery Data Service*

1. Imagery innovation–research-ready imagery-based data products, building off and developing innovative computing and AI methods that facilitate efficient automated workflows for measures and indicators, as well as custom-defined geographies and time periods.

2. Data for all–data distribution channels that meet researchers and policymakers where they are, with user-friendly interfaces, familiar file formats, linkage and integration with existing data resources

3. Capability and community–building capacity for understanding and working with imagery and imagery-derived data, growing the user-base and providing thought leadership, and heightening awareness and enthusiasm for the value of imagery


Epilogue

Complex spatial inequality challenges require complex approaches1

  • Inter-disciplinarity
  • Geography
  • A lot of social science under the hood 2

The question chooses the method.

(Not the other way around)

Last takeaways

(through the lens of yesterday’s conversation)

  • We really mustn’t forget to talk about data
  • Basic principles of good estimators (i.e., methods): consistent, unbiased, efficient
  • How does GeoAI (however we are defining it) fit into the broader suite of existing geographical methods? Do we draw boundaries? Build walls? Moats?
  • I’d love to see us all talking to each other more

The End.